US9165113B2  System and method for quantitative assessment of frailty  Google Patents
System and method for quantitative assessment of frailty Download PDFInfo
 Publication number
 US9165113B2 US9165113B2 US13283337 US201113283337A US9165113B2 US 9165113 B2 US9165113 B2 US 9165113B2 US 13283337 US13283337 US 13283337 US 201113283337 A US201113283337 A US 201113283337A US 9165113 B2 US9165113 B2 US 9165113B2
 Authority
 US
 Grant status
 Grant
 Patent type
 Prior art keywords
 frailty
 inertial sensor
 sensor data
 set
 reference
 Prior art date
 Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
 Active, expires
Links
Images
Classifications

 G06F19/3431—

 A—HUMAN NECESSITIES
 A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
 A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
 A61B5/00—Detecting, measuring or recording for diagnostic purposes; Identification of persons
 A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
 A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
 A61B5/1116—Determining posture transitions
 A61B5/1117—Fall detection

 G—PHYSICS
 G06—COMPUTING; CALCULATING; COUNTING
 G06F—ELECTRIC DIGITAL DATA PROCESSING
 G06F19/00—Digital computing or data processing equipment or methods, specially adapted for specific applications

 G—PHYSICS
 G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
 G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
 G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
 G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment

 A—HUMAN NECESSITIES
 A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
 A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
 A61B2562/00—Details of sensors; Constructional details of sensor housings or probes; Accessories for sensors
 A61B2562/02—Details of sensors specially adapted for invivo measurements
 A61B2562/0219—Inertial sensors, e.g. accelerometers, gyroscopes, tilt switches
Abstract
Description
Embodiments generally relate to frailty assessment, and more particularly to quantifying frailty based on data collected by bodyworn sensors.
The concept of frailty in elderly adults has been gaining acceptance as an emerging geriatric syndrome. Although frailty is a recognizable and common phenomenon in aging, it can be a concept that is difficult to accurately define and diagnose. Frailty is a multifactorial condition, influenced by the combination of a person's physical, psychological, and social health. Frailty has been shown to have a predictive association with important health outcomes such as first fall, first hospitalization, worsening ADL disability, worsening mobility disability, and death.
Accurate assessment of a patient's frailty level could allow for effective multifactorial intervention. Frailty has previously been assessed using subjective clinical judgment. More recently, validated clinical scales, such as the Fried frailty index, has been used to quantify frailty levels. The Fried frailty index ranges from 0 to 5. A value of 0 is assigned to individuals considered robust. A value of 12 is assigned to individuals considered prefrail, and a value of 35 is assigned to individuals considered frail. Assessing an individual's Fried frailty index requires, however, statistical expertise and a reference sample. The assessment further requires recognition of unintentional weight loss, weakness (e.g., low grip strength), slow walking speed, low physical activity, and/or selfmonitoring for exhaustion. Such expertise and resources are often not available in a primary care setting, where primary care practitioners are often nonexperts and are not trained to measure or assess clinical metrics such as the Fried frailty index.
One aspect of the invention relates to estimating an individual's Fried frailty index based on data collected by one or more bodyworn inertial sensors. The inertial sensor data may be collected during a walking trial, such as a timedupandgo (TUG) test. Parameters quantified by the inertial sensor data may be used as input parameters in a model (e.g., a regression model) that assesses the individual's frailty. The model may include all input parameters that have or exceed a threshold level of significance (e.g., P≧0.05) to the output frailty value. The model may be configured to output frailty values that approximate clinicallymeasured frailty index values. For example, a linear regression model may be generated to approximate reference, clinicallymeasured frailty index values of individuals. The output frailty values may be estimated from inertial sensor data generated by the individuals. The linear regression model may be configured to approximate the frailty index value with a high coefficient of determination, or R^{2 }value. A high R^{2}, which can have a maximum value of 1, indicates that the regression model explains a high degree of any patterns and variations in the reference frailty index values. Other measures of a regression model's accuracy, such as sensitivity or specificity, may also be used. They generally measure how often a regression model outputs a frailty classification that matches a reference clinical classification, and are discussed more below.
The regression models may be used to subsequently quantify frailty based on an individual's inertial sensor data. The calculation would no longer require specialized equipment, specialized expertise, or access to a reference sample. Nonexperts in a primary care setting may thus be able to use the models to generate a quantitative frailty estimate of their patients. In one example, a linear regression model yielded a R^{2 }value of about 0.39, showing that the frailty quantified by the regression model may also approximate a clinicallymeasured frailty index value with an accuracy corresponding to R^{2}=0.39.
In some implementations, one or more regression models may output a frailty class rather than a frailty index value. For example, a regression model may output an approximation of whether an individual is frail, prefrail, or robust based on his or her inertial sensor data. In a more specific example, two logistic regression models may be generated to classify an individual's frailty. Each of the logistic regression models may compare the individual's inertial sensor data against the inertial sensor data collected from individuals who were classified as robust. A first logistic regression model may distinguish between whether the individual is frail or robust. A second logistic regression model may distinguish between whether the individual is prefrail or robust. The accuracy of the regression models may be measured based on their sensitivity and specificity, which are defined below. In one example, a first logistic regression model yielded an accuracy of about 88%. The first logistic regression model may thus be used to distinguish between frail and robust individuals in a way that closely approximates clinicallybased distinctions of frail and robust individuals. In the example, a second logistic regression model yielded an accuracy of 58%. The second logistic regression model may thus be used when more granularity of an individual's frailty class is desired (i.e., whether the individual is prefrail, a category between frail and robust) and the accuracy in approximating clinicallydetermined frailty classes is not as important.
The regression models may thus be used to quantify an individual's frailty and to classify the individual's frailty into one of a plurality of frailty classes. The regression models may be implemented by any device, such as software running on a processor, that is configured to receive values of inertial sensor parameters, which are used as inputs in the regression models. The implementation thus provides a quantitative assessment of frailty that may be administered by nonexperts in a clinical setting, a primary care setting, a home setting, or any other setting.
The various advantages of the embodiments of the present invention will become apparent to one skilled in the art by reading the following specification and appended claims, and by referencing the following drawings, in which:
Embodiments may provide for a system that generates inertial sensor data from an individual's movement and uses the data to quantify the individual's frailty. The quantified frailty may approximate clinical measures of the individual's frailty, such as the Fried frailty index.
The individual may generate the inertial sensor data during a timed up and go (TUG) test.
Sensors 14 a, 14 b may measure various inertial sensor parameters related to the person's motion. Some parameters include angular velocity parameters, acceleration parameters, spatial gait parameters, and temporal gait parameters. Characteristic points, such as an initial contact with the ground (e.g., a heel strike), a terminal contact with the ground (e.g., a toeoff), and a midswing point may be identified from the parameters' values.
Other inertial sensor parameters, such as the time taken to perform the TUG test, may be recorded by a clinician or primary care practitioner administering the test, or may be calculated from the angular velocity data. The time parameters may include the manual TUG time, which begins when the individual 10 is instructed to begin the test and ends when the individual 10 sits back on the chair. The time parameters include a turn time, which identifies the amount of time between the individual 10's first step and the step in which the individual 10 turns from spot 12. The time parameters may further include a return time, which identifies the amount of time between the turn step and the last step taken by the individual 10 in the TUG test.
The sensors may include triaxial accelerometers, triaxial gyroscopes, GPS transceivers, passive infrared (PIR) sensors, tilt and vibration sensors, and any other sensor operable to measure movement or forcerelated parameters. Each sensor may be configured to measure rotation and acceleration about a X, Y, and Z sensor axis, as illustrated in
At operation 215, the inertial sensor data may be calibrated. For example, the raw gyroscope and accelerometer data may be calibrated to derive the angular velocity and acceleration vectors with respect to the sensors' unit coordinate axes or some other orientation. In the example, operation 215 may use the sensors' gyroscope data to provide a sensortosegment offset orientation matrix (e.g., a rotation matrix) to calibrate the data to derive acceleration and angular velocity vectors with respect to the coordinate axes of each inertial sensor. Any other standard calibration procedure may be used to calibrate the gyroscopes before they are used in the walking trial.
At operation 220, the calibrated inertial sensor data may be filtered before or after transmission from the sensor to remove noise. For example, a zerophase 5^{th }order Butterworth filter may be used to low pass filter the inertial sensor data. The corner frequency (e.g., 50.2 Hz) may be calculated as
where f_{s }is the sampling rate. In the example, a low pass filter may be applied to data that measures angular velocity in the mediolateral direction. A bandpass filter may also be used to filter out lowfrequency components of the data.
At operation 225, adaptive thresholds may be calculated based on the angular velocity data. The adaptive thresholds may be used to define the likely range of characteristic points, namely initial contact points, terminal contact points, and midswing points. Such points are located generally in local maxima or minima of the angular velocity data. Restricting examination of the angular velocity data to portions isolated by the adaptive thresholds can ensure robust detection of the characteristic points over a variety of walking speeds.
For example, angular velocity data and identified characteristic points are shown in
th _{1}=0.6·max(ω_{ML}), (1)
where ω_{ML }is the medio lateral angular velocity. The local maximum, midswing point may further be verified to have an angular velocity above a second adaptive threshold th_{2}, where
Moreover, if two maximum peaks are found within t_{1 }seconds of each other, only the greater maximum can be considered, wherein t_{1 }may be defined as 0.5 seconds or f_{s}*1.5 and f_{s }is defined as the stride frequency.
The adaptive thresholds may further be used to identify, at operation 230, initial contact points (e.g., heel strike points) and terminal contact points (e.g., toeoff points).
The local minimum for the initial contact point may further be required to be less than adaptive threshold th_{5 }rad/sec, where
th _{5}=mean(ω_{ML}) (4)
A terminal contact point may be located at a local minimum that has an angular velocity less than adaptive threshold th_{4 }and that has a preceding maximum that is greater than adaptive threshold th_{6}, where
In identifying initial contact and terminal contact points, only data within t_{2 }seconds may be considered. In one example t_{2 }may be defined as 1.5 seconds for f_{s}*1.5. Specific values and ranges are provided as examples only, and other values and ranges may be used as appropriate.
At operation 235, spatiotemporal parameters of the individual 10's measured gait may be calculated. The parameters may include measurements such as the individual 10's cadence, number of gait cycles, number of steps taken, stance time, step time, swing time, single support percentage, double support percentage, and stride time, that may be included in a regression model. The parameters may further include statistical derivations of the measurements, such as such as their mean, maximum, range, and coefficient of variation (CV).
The number of gait cycles may be calculated as the number of initial contact (e.g., heel strike) points minus one, and may represent the number of complete gait cycles. The cadence may measure the number of steps per minute, and may be calculated as the number of steps divided by the TUG walking time (in units of minutes). The step time may be calculated as the time between the initialcontact (e.g., heelstrike) point on one foot and the initial contact point on the other foot. The stride time may be calculated as the time from initial contact (e.g., heelstrike) of one foot to a subsequent initial contact of the same foot. Stance time may be calculated as the time from an initial contact (e.g., heelstrike) point to a terminal contact (e.g., toeoff) point on the same foot. Swing time may be calculated as the time from a terminal contact (e.g., toeoff) point to a initial contact (e.g., heelstrike) point on the same foot. Double support percentage may be determined by calculating the percentage of each gait cycle during which both feet are in contact with the ground (where the gait cycle can be calculated as the time between successive initial contact points (e.g., heelstrike points)). Single support percentage for a foot may be defined as the swing duration of the other foot expressed as a percentage of gait cycle time.
The angular velocity data may be used to determine inertial sensor parameters like TUG time segments, which include a walk time, turn time, and return time.
Operation 240 demonstrates that in addition to the spatiotemporal gait parameters, the derived parameters may include one or more parameters that are obtained directly from the angular velocity data. The angular velocity data may be divided among along the sensor's Y, X, and Z axes. The axes may correspond to an individual's mediolateral (ML), anteroposterior (AP), and V directions. These angular velocity parameters could include parameters to detect and analyze the speed and timing of an individual's turn during the TUG test. For example, the mean, minimum and maximum angular velocities (averaged across both shanks) during the walk, expressed in degrees per second, may each be determined in the Y, X, and Z axes (which may correspond to, for example, the ML, AP and V directions). The measurements may form a set of nine (i.e., 3×3) triaxial angular velocity parameters. Further, the coefficient of variation (CV) of the parameters may also be calculated. The CV of a parameter may be defined as the ratio of the parameter's standard deviation to its mean.
The triaxial set of angular velocities may also be multiplied by the height of the individual performing the TUG test in order to obtain a variable approximately proportional to the linear velocity of the shank. This approximation can be based on the formula for linear velocity, which equals the radius times angular velocity, wherein the radius is the leg length and height is assumed to be approximately proportional to the leg length. Thus, the linear velocity may be specifically related to the shank/foot of the individual as opposed to merely the trunk of the individual. In addition, other angular velocitybased parameters such as turn angular velocity may be calculated. The turn angular velocity can be defined as the mean amplitude (taken across both shanks) of the angular velocity signal at the turn point for each shank.
At operation 245, the individual 10's eyesight may also be measured, such as on a Binocular logmar or a PelliRobson contrast sensitivity scale.
At operation 250, artefact rejection may be performed. The artefact rejection may be configured to remove spurious temporal parameters that might have been calculated from erroneous gyroscope data. The artefact rejection routine may be designed to account for missing and extra initial contact and terminal contact points. Further, the artefact rejection may be based on two strands: examining temporal sequence information, and examining times between successive characteristic points (e.g., gait cycle information).
Temporal sequence information may be obtained based on the following: once all characteristic points (e.g., initial contact points, terminal contact points) are detected, each point may be assigned a numerical label of one to four. For example, an initial contact point on the right foot is assigned 1; a terminal contact point on the left foot is assigned 2; an initial contact point on the left foot is assigned 3; and a terminal contact point on the right foot is assigned 4. A correct gait cycle (if starting on a right initial contact point) would then follow the sequence 1, 2, 3, 4. By subtracting each label from the previous label, spurious samples (e.g., samples not producing a difference equal to either −3 or 1) may be deemed artefacts and rejected.
The time between successive characteristic points (e.g., gait cycle information) may be calculated for each set of four left and right initial contact and terminal contact points. This calculation may be referred to as “gait cycle time.” If the difference between any successive characteristic point is greater than a particular time threshold (e.g., 2.5 seconds), the associated characteristic point could be identified as an artefact. Similarly, if the difference between any successive characteristic point is zero seconds, the associated point may be flagged as an artefact. Furthermore, any gait parameters with a negative or zero value may also be rejected.
The collection and processing of inertial sensor and other data in operations 210 to 250 are described in more detail in copending applications entitled “WIRELESS SENSOR BASED QUANTITATIVE FALLS RISK ASSESSMENT,” U.S. application Ser. No. 12/782,110; and “A METHOD FOR BODYWORN SENSOR BASED PROSPECTIVE EVALUATION OF FALLS RISK IN COMMUNITYDWELLING ADULTS,” U.S. application Ser. No. 13/186,709, both of which are incorporated herein by reference.
At operation 260, the processed values of the inertial sensor parameters may be used to generate a quantitative frailty estimate. The quantitative frailty estimate may approximate a clinicallymeasured frailty metric. The metric may be a Fried frailty index value, a frailty class based on the index value, or any other frailty metric. In one example, a linear regression model may quantify an individual's frailty. The quantity outputted by the linear regression model may further be an estimate of what a clinicallymeasured frailty index value for the individual would have been. In one example, a logistic regression model may classify the individual's frailty level as prefrail, frail, or robust. The class outputted by the logistic regression model may correspond to the classes defined by Fried, and may estimate what a clinicallymeasured classification for the individual would have been. The regression models may output their estimates based on inertial sensor data generated by the individual.
The regression models may initially be generated from reference Fried frailty index values and classifications that were obtained, for example, in a clinical setting using the Fried standard method. The Fried standard method may determine an index value based on three or more of the following five criteria: unintentional weight loss, selfreported exhaustion, weakness (e.g., lowered grip strength), slow walking speed, and low physical activity. The frailty index value may quantify the combination of the frailty criteria as number from 0 to 5. An index value of 0 classifies the individual as robust. An index value of 12 classifies the individual as prefrail. An index value of 35 classifies the individual as frail.
The frailty index values may be used to generate a regression model that approximates the index values using inertial sensor data. The regression model may be used to subsequently estimate an individual's frailty index value or frailty class from the individual's inertial sensor data. By using the generated regression model, the individual's frailty may later be quantified without a separate analysis of Fried's five frailty criteria and without the need for personnel having statistical expertise. Rather, the individual's frailty may be quantified with one or more inertial sensors attached to the individual and a processor that implements the regression model. The quantified frailty may be treated as an estimate of a clinicallymeasured frailty metric.
In some implementations, a linear regression model may be generated at operation 260 to estimate a frailty index value. The input to the linear regression model may comprise any inertial sensor parameter quantified by the inertial sensor data. If the data generates a large quantity of parameters, the number of parameters included in the model may be reduced. For example, all the inertial sensor parameters may be grouped into blocks in terms of general characteristics. To generate the linear regression model, a series of linear regression analyses may be performed on the parameters and on all twoway interactions of parameters in each block. The linear regression analyses may determine the significance (e.g., pvalue) of any relationship between the reference frailty index values and each inertial sensor parameter or twoway interaction. In the linear regression analyses, the reference frailty index value may be treated as the dependent variable while the inertial sensor parameter or twoway interaction may be treated as an independent variable. Independent variables that have insufficient significance to the dependent variable, such as those having p<0.05, may be excluded from the linear regression model being generated. The remaining variables, or parameters, from each block may be combined into a final linear regression model. An example of inertial sensor parameters and twoway interactions that are included in a linear regression model is shown in Table 1. The inertial sensor parameters and twoway interactions form the input that is used to estimate a frailty index value.
TABLE 1  
Mean single support  Mean AP Ang. vel.  
(%)  
Cadence  Min AP Ang. vel.  
Mean ML ang. Vel  No steps  
Max ML ang vel.  Mean single support:Cadence  
Min ML ang. vel.  Max ML ang. vel.:Min ML  
ang. vel.  
Time of turn  Time of turn:Turn ang. vel.  
ML ang. vel. at turn  Mean AP ang. Vel.:Min AP  
ang. vel.  
Max AP ang. vel.  Max AP ang. vel.:No steps  
The association of each quantitative parameter with the Fried frailty index may further be examined using Pearson's correlation coefficient. Separate coefficients may be calculated for male individuals and female individuals. Table 2 shows example results of bivariate correlations that yielded the absolute value of Pearson's correlation coefficient between each of a group of inertial sensor parameters and reference frailty index values. Parameters in the table are ranked according to the absolute value of the coefficient.
TABLE 2  
Absolute value of  
Pearson's Correlation  
Variable  Coefficient 
Walk time  0.55 
Time of turn  0.55 
Return time  0.55 
Number of steps  0.49 
Mean time to turn  0.46 
Gait cycles  0.45 
TUG recording time  0.40 
Mean midswing points  0.39 
Max mediolateral angular velocity × height  0.37 
Min AP angular velocity  0.36 
Mean stride velocity  0.36 
Max mediolateral angular velocity  0.34 
Min mediolateral angular velocity × height  0.34 
Mean stride length  0.33 
Min mediolateral angular velocity  0.31 
Cadence  0.29 
Min vertical angular velocity × height  0.29 
Min AP angular velocity × height  0.29 
Max AP angular velocity × height  0.27 
Max vertical angular velocity  0.27 
Max AP angular velocity  0.26 
Mean AP angular velocity  0.26 
Coefficient of variation of stride length  0.25 
Mean stance time  0.24 
Max vertical × height  0.23 
Mean mediolateral angular velocity × height  0.23 
Min vertical angular velocity  0.22 
Mean vertical angular velocity  0.22 
Mean single support  0.21 
CV stride velocity  0.21 
Mean ML angular velocity  0.21 
Range of midswing points  0.20 
Mean turn ratio  0.20 
Mean stride time  0.15 
The performance of the linear regression model may be measured with a R^{2 }value, which determines how well the model explains the variation in the output. For example, the R^{2 }may measure whether variations in reference frailty index values can be explained by a linear combination values of inertial sensor parameters. A R^{2 }value equal to zero implies that none of the variation is explained while a unity value means that all the variation in the output, reference index values is explained by the linear regression model. One example linear regression model yielded a R^{2 }value of 0.3873, suggesting that approximately 39% of variation in reference frailty index values is explained by the model.
In addition to or instead of estimating an individual's frailty index value, the individual's frailty class (e.g., prefrail, frail, or robust) may be estimated. In one example, the frailty class may be approximated through two separate models. A first model may be based on dichotomous comparisons of individuals clinically classified as frail against individuals clinically classified as robust, in which the latter may serve as a reference class. A second model may be based on dichotomous comparisons of individuals clinically classified as prefrail against the individuals in the reference class. The comparisons may be based on the inertial sensor data generated by the individuals in one class versus inertial sensor data generated by individuals in the reference class. The first model may estimate whether the individual is in the prefrail class or is in the robust class. The second model may estimate whether the individual is in the frail class or is in the robust class. In one example, the regression model may be a logistic regression model.
The input to the logistic or any other regression model may comprise any inertial sensor parameter quantified by the inertial sensor data. Like with the linear regression model, the number of input parameters may be reduced by grouping the parameters into blocks and performing a logistic regression analysis on all parameters and on all twoway interactions between parameters in each block. The regression analyses may identify parameters that do not have a significant relationship (e.g., p<0.05) with the logistic model's dependent parameter, which may comprise frailty classes that were clinically measured. The inertial sensor parameters that are not significant may be excluded from the final logistic regression model, while the significant parameters may be included. In order to obtain unbiased generalized models, each model was validated using 1000 random subsampling cross validation. Table 3 shows inertial sensor parameters that are included in an example logistic regression model that distinguishes between prefrail versus robust individuals and in an example logistic regression model that distinguishes between frail versus robust individuals.
TABLE 3  
Model 1  Model 2  
PreFrail Vs Robust  Frail Vs Robust  
Mean double Support  Mean single support  
Mean single Support  Cadence  
CV double Support  Max ML ang. vel.  
CV stride Time  Mean single support:  
Cadence  
Min AP ang. vel.  
Mean AP Angle Velocity  
Mean double support:Mean  
single support  
CV double support:CV stride  
time  
Min AP ang. vel.:Mean AP  
ang. vel.  
Multinomial regression analysis may be used to compare how well the first and second models distinguish against individuals in a robust group. For example, one multinomial regression analysis was performed with a group of 110 individuals clinically classified as prefrail, 102 individuals clinically classified as robust, and 23 individuals clinically classified as frail. The analysis in the example showed that prefrail patients may tend to be less different when compared with robust individuals than when compared with frail individuals. Table 4 below shows that the class of frail individuals in one example had more parameters that were significantly different (as shown by low Pvalues) from the robust class.
TABLE 4  
PreFrail (N = 110)  Frail (N = 23) 
Walk time (p = 0.001)  Mean single support (p < .001) 
Gait cycles (p = 0.009)  Mean stance time (p < .001) 
No steps (p = 0.01)  Mean stride time (p = 0.003) 
Return time (p < 0.001)  CV stride time (p = 0.03) 
Time of turn (p = 0.007)  CV step time (p = 0.013) 
Min AP ang. Vel. (p = 0.013)  Mean ML Ang. Vel. (p < .001) 
Mean stride velocity (p = 0.001)  Max ML Ang. Vel. (p < .001) 
Means stride length (p = 0.002)  Min ML Ang. Vel. (p < .001) 
CV stride length (p = 0.01)  Walk time (p < 0.001) 
Mean turning time (p = 0.026)  Gait cycles (p < 0.001) 
No steps (p < 0.001)  
Cadence (p < 0.001)  
Walk ratio (p = 0.013  
Return time (p < 0.001)  
Range of mid swing points (p = 0.002)  
Mean midswing points (p <0.001)  
Time of turn (p < 0.001)  
TUG recording time (p < 0.001)  
Mean AP Ang. Vel. (p < 0.001)  
Max AP Ang. Vel. (p < 0.001)  
Min AP Ang. Vel. (p < 0.001)  
Mean V. Ang. Vel. (p = 0.002)  
Max V. Ang. Vel. (p < 0.001)  
Min V. Ang. Vel. (p = 0.001)  
Mean stride velocity (p < 0.001)  
CV stride velocity (p = 0.005)  
Mean stride length (p < 0.001)  
CV stride length (p < 0.001)  
Mean turning time (p < 0.001)  
Mean turn ratio (p < 0.001)  
Mean ML ang. Vel. × Height  
(p < 0.001)  
Max ML ang. Vel. × Height (p < 0.001)  
Min ML ang. Vel. × Height (p < 0.001)  
Max AP ang. Vel. × Height (p < 0.001)  
The performance of the logistic regression model in classifying individuals into a frailty class may be quantified using sensitivity (“sens”) and specificity (“spec”). Sensitivity may be calculated as the proportion of prefrail or frail individuals (e.g., individuals having a Fried frailty index value of 12 and 35, respectively) “correctly” identified by the logistic regression model. An identification by the logistic regression model may be considered correct if it matches a reference, clinicallymeasured classification. Specificity may be calculated as the proportion of robust individuals (e.g., individuals having a frailty index value of 0) “correctly” identified by the logistic regression model. Table 5 shows example sensitivity and specificity values for the two logistic regression models discussed above. The table shows that the model distinguishing between prefrail and robust individuals has an average accuracy of 57.9%, while the model distinguishing between frail and robust individuals has an average accuracy of 87.2%.
TABLE 5  
Model 1  Model 2  
Sens (%)  61.3  87.8  
Spec (%)  54.4  86.8  
The logistic regression model that distinguishes between robust and frail individuals, such as Model 2, may thus be used to distinguish between frail and robust individuals in a way that closely approximates clinicallybased distinctions of frail and robust individuals. The logistic regression model that distinguishes between robust and prefrail individuals, such as Model 1, may not have as high a degree of accuracy, but provides a more granular determination of an individual's frailty level.
Generation of regression models is discussed in more detail in copending application “WIRELESS SENSOR BASED QUANTITATIVE FALLS RISK ASSESSMENT,” U.S. application Ser. No. 12/782,110, which is incorporated herein by reference.
The operations described above may be implemented in executable software as a set of logic instructions stored in a machine or computerreadable medium of a memory such as random access memory (RAM), read only memory (ROM), programmable ROM (PROM), firmware, flash memory, etc., in fixedfunctionality hardware using circuit technology such as application specific integrated circuit (ASIC), complementary metal oxide semiconductor (CMOS) or transistortransistor logic (TTL) technology, or any combination thereof. For example, computer program code to carry out operations may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like and conventional procedural programming languages, such as the “C” programming language or similar programming languages.
Embodiments of the present invention are applicable for use with all types of semiconductor integrated circuit (“IC”) chips. Examples of these IC chips include but are not limited to processors, controllers, chipset components, programmable logic arrays (PLA), memory chips, network chips, and the like. In addition, in some of the drawings, signal conductor lines are represented with lines. Some may be thicker, to indicate more constituent signal paths, have a number label, to indicate a number of constituent signal paths, and/or have arrows at one or more ends, to indicate primary information flow direction. This, however, should not be construed in a limiting manner. Rather, such added detail may be used in connection with one or more exemplary embodiments to facilitate easier understanding of a circuit. Any represented signal lines, whether or not having additional information, may actually comprise one or more signals that may travel in multiple directions and may be implemented with any suitable type of signal scheme, e.g., digital or analog lines implemented with differential pairs, optical fiber lines, and/or singleended lines.
Example sizes/models/values/ranges may have been given, although embodiments of the present invention are not limited to the same. As manufacturing techniques (e.g., photolithography) mature over time, it is expected that devices of smaller size could be manufactured. In addition, well known power/ground connections to IC chips and other components may or may not be shown within the figures, for simplicity of illustration and discussion, and so as not to obscure certain aspects of the embodiments of the invention. Further, arrangements may be shown in block diagram form in order to avoid obscuring embodiments of the invention, and also in view of the fact that specifics with respect to implementation of such block diagram arrangements are highly dependent upon the platform within which the embodiment is to be implemented, i.e., such specifics should be well within purview of one skilled in the art. Where specific details (e.g., circuits) are set forth in order to describe example embodiments of the invention, it should be apparent to one skilled in the art that embodiments of the invention can be practiced without, or with variation of, these specific details. The description is thus to be regarded as illustrative instead of limiting.
Those skilled in the art will appreciate from the foregoing description that the broad techniques of the embodiments of the present invention can be implemented in a variety of forms. Therefore, while the embodiments of this invention have been described in connection with particular examples thereof, the true scope of the embodiments of the invention should not be so limited since other modifications will become apparent to the skilled practitioner upon a study of the drawings, specification, and following claims.
Claims (8)
Priority Applications (1)
Application Number  Priority Date  Filing Date  Title 

US13283337 US9165113B2 (en)  20111027  20111027  System and method for quantitative assessment of frailty 
Applications Claiming Priority (1)
Application Number  Priority Date  Filing Date  Title 

US13283337 US9165113B2 (en)  20111027  20111027  System and method for quantitative assessment of frailty 
Publications (2)
Publication Number  Publication Date 

US20130110475A1 true US20130110475A1 (en)  20130502 
US9165113B2 true US9165113B2 (en)  20151020 
Family
ID=48173268
Family Applications (1)
Application Number  Title  Priority Date  Filing Date 

US13283337 Active 20331221 US9165113B2 (en)  20111027  20111027  System and method for quantitative assessment of frailty 
Country Status (1)
Country  Link 

US (1)  US9165113B2 (en) 
Families Citing this family (2)
Publication number  Priority date  Publication date  Assignee  Title 

JP6152763B2 (en) *  20130919  20170628  カシオ計算機株式会社  Exercise support apparatus and exercise support method, exercise support program 
EP3218831A1 (en) *  20141112  20170920  Koninklijke Philips N.V.  Method and apparatus for quantifying and monitoring the frailty of a subject 
Citations (17)
Publication number  Priority date  Publication date  Assignee  Title 

US6820025B2 (en) *  20001030  20041116  The United States Of America As Represented By The Secretary Of The Navy  Method and apparatus for motion tracking of an articulated rigid body 
US20080146968A1 (en) *  20061214  20080619  Masuo Hanawaka  Gait analysis system 
US20090076765A1 (en) *  20060303  20090319  Garmin Ltd.  Method and apparatus for estimating a motion parameter 
US20090185772A1 (en) *  20080122  20090723  General Electric Company  Fiberoptic patient health multiparameter monitoring devices and system 
US20090247910A1 (en) *  20040107  20091001  David Klapper  Method and apparatus for classification of movement states in Parkinson's disease 
US20110022349A1 (en) *  20060303  20110127  Garmin Switzerland Gmbh  Method and apparatus for estimating a motion parameter 
US20110162433A1 (en) *  20080912  20110707  Koninklijke Philips Electronics N.V.  Fall detection system 
US20110213278A1 (en) *  20100226  20110901  Apdm, Inc.  Movement monitoring system and apparatus for objective assessment of movement disorders 
US20110288811A1 (en) *  20100518  20111124  Greene Barry R  Wireless sensor based quantitative falls risk assessment 
US20120065915A1 (en) *  20100913  20120315  Fujitsu Limited  Information processing method, apparatus, and program 
US20120072168A1 (en) *  20090520  20120322  Koninklijke Philips Electronics N.V.  Sensing device for detecting a wearing position 
US20120253234A1 (en) *  20090903  20121004  Ming Young Biomedical Corp.  System and method for analyzing gait using fabric sensors 
US20120289791A1 (en) *  20110513  20121115  Fujitsu Limited  Calculating and Monitoring the Efficacy of StressRelated Therapies 
US20120316843A1 (en) *  20110608  20121213  Cobra Golf Incorporated  Systems and methods for communicating sportsrelated information 
US20130060512A1 (en) *  20110901  20130307  IntelGe Care Innovations Llc  Calculation of minimum ground clearance using body worn sensors 
US20130123669A1 (en) *  20100727  20130516  Omron Healthcare Co Ltd  Gait change determination device 
US20130123666A1 (en) *  20050317  20130516  Great Lakes Neurotechnologies Inc.  Movement disorder recovery system and method for continuous monitoring 
Patent Citations (18)
Publication number  Priority date  Publication date  Assignee  Title 

US6820025B2 (en) *  20001030  20041116  The United States Of America As Represented By The Secretary Of The Navy  Method and apparatus for motion tracking of an articulated rigid body 
US20090247910A1 (en) *  20040107  20091001  David Klapper  Method and apparatus for classification of movement states in Parkinson's disease 
US20130123666A1 (en) *  20050317  20130516  Great Lakes Neurotechnologies Inc.  Movement disorder recovery system and method for continuous monitoring 
US20090076765A1 (en) *  20060303  20090319  Garmin Ltd.  Method and apparatus for estimating a motion parameter 
US20110022349A1 (en) *  20060303  20110127  Garmin Switzerland Gmbh  Method and apparatus for estimating a motion parameter 
US20080146968A1 (en) *  20061214  20080619  Masuo Hanawaka  Gait analysis system 
US20090185772A1 (en) *  20080122  20090723  General Electric Company  Fiberoptic patient health multiparameter monitoring devices and system 
US20110162433A1 (en) *  20080912  20110707  Koninklijke Philips Electronics N.V.  Fall detection system 
US20120072168A1 (en) *  20090520  20120322  Koninklijke Philips Electronics N.V.  Sensing device for detecting a wearing position 
US20120253234A1 (en) *  20090903  20121004  Ming Young Biomedical Corp.  System and method for analyzing gait using fabric sensors 
US8961439B2 (en) *  20090903  20150224  Ming Young Biomedical Corp.  System and method for analyzing gait using fabric sensors 
US20110213278A1 (en) *  20100226  20110901  Apdm, Inc.  Movement monitoring system and apparatus for objective assessment of movement disorders 
US20110288811A1 (en) *  20100518  20111124  Greene Barry R  Wireless sensor based quantitative falls risk assessment 
US20130123669A1 (en) *  20100727  20130516  Omron Healthcare Co Ltd  Gait change determination device 
US20120065915A1 (en) *  20100913  20120315  Fujitsu Limited  Information processing method, apparatus, and program 
US20120289791A1 (en) *  20110513  20121115  Fujitsu Limited  Calculating and Monitoring the Efficacy of StressRelated Therapies 
US20120316843A1 (en) *  20110608  20121213  Cobra Golf Incorporated  Systems and methods for communicating sportsrelated information 
US20130060512A1 (en) *  20110901  20130307  IntelGe Care Innovations Llc  Calculation of minimum ground clearance using body worn sensors 
NonPatent Citations (12)
Title 

A. Salarian, F. B. Horak, C. Zampieri, P. C. Kuhta, J. G. Nutt, and K. Aminian, "iTUG, a Sensitive and Reliable Measure of Mobility", pp. 303310, Jun. 2010. * 
B. Greene, D. McGrath, R. O'Neill, K. J. O'Donovan, A. Burns, B. Caulfield, An Adaptive gyroscopebased algorithm for temporal gait analysis pp. 12511260, May 2010. * 
B. R. Greene, A. O'Donovan, R. R. Ortuno, L. Cogan, C. N. Scanaill, and R. A. Kenny, "Quantitative Falls Risk Assessment Using the Timed Up and Go Test", pp. 29182926, Dec. 2010. * 
B. R. Greene, D. McGrath, K. J. O'Donovan, R. O'Neill, A. Burns, B. Caulfield, "Adaptive estimation of temporal gait parameters using bodyworn gyroscopes" pp. 12961299, 2010. * 
D. McGrath, B. R. Greene, C. Walsh, B. Caulfield, "Estimation of minimum ground clearance (MGC) using bodyworn inertial sensors", pp. 10831088, 2011. * 
E. P. Doheny, T. G. Foran, and B. R. Greene, "A single gyroscope method for spatial gait analysis" Sep. 4, 2010, pp. 13001303. * 
Fried, et al. "Frailty in Older Adults: Evidence for a Phenotype" Journal of Gerantology: Medical Sciences, 2001, vol. 56A, No. 3, pp. M146M156. 
K. O'Donovan, B. R. Greene, D. McGrath, R. O'Neill, A. Burns, B. Caulfield, Shimmer: a new tool for temporal Gait analysis, Sep. 2009, pp. 38263829. * 
M. A. Makary, D. L. Segev, P. J. Pronovost, D. Syin, K. BandeenRoche, P. Patel, R. Takenaga, L. Devgan, C. G. Holzmueller, J. Tian, L. P. Fried, "Frailty as a Predictor of Surgical Outcomes in Older Patients" pp. 901908, 2010. * 
M. R. Narayanan, S. J. Redmond, M. E. Sealzi, S. R. Lord, B. G. Celler, N. H. Lovell, "Longitudinal Falls Risk Estimation using Triaxial Accelerometry" pp. 18, 2009. * 
Y. Higashi, K. Yamakoshi, T. Fujimoto, M. Sekine, and T. Tamura, "Quantitative Evaluation of Movement using the timed UpandGo Test" pp. 3846, 2008 IEEE. * 
Y. Higashi, K. Yamakoshi, T. Fujimoto, M. Sekine, and T. Tamura, "Quantitative Evaluation of Movement Using the Timed UpandGo test". pp. 3846, 2008. * 
Also Published As
Publication number  Publication date  Type 

US20130110475A1 (en)  20130502  application 
Similar Documents
Publication  Publication Date  Title 

Preece et al.  A comparison of feature extraction methods for the classification of dynamic activities from accelerometer data  
Karantonis et al.  Implementation of a realtime human movement classifier using a triaxial accelerometer for ambulatory monitoring  
Salarian et al.  Ambulatory monitoring of physical activities in patients with Parkinson's disease  
Lyden et al.  A comprehensive evaluation of commonly used accelerometer energy expenditure and MET prediction equations  
Atallah et al.  Sensor positioning for activity recognition using wearable accelerometers  
Greene et al.  Quantitative falls risk assessment using the timed up and go test  
Rosenberger et al.  Estimating activity and sedentary behavior from an accelerometer on the hip or wrist  
Howcroft et al.  Review of fall risk assessment in geriatric populations using inertial sensors  
Bugané et al.  Estimation of spatialtemporal gait parameters in level walking based on a single accelerometer: Validation on normal subjects by standard gait analysis  
US8206325B1 (en)  Ambulatory system for measuring and monitoring physical activity and risk of falling and for automatic fall detection  
US8529448B2 (en)  Computerized systems and methods for stability—theoretic prediction and prevention of falls  
Altini et al.  Estimating energy expenditure using bodyworn accelerometers: a comparison of methods, sensors number and positioning  
Storm et al.  Step detection and activity recognition accuracy of seven physical activity monitors  
Crouter et al.  Validity of ActiGraph 2regression model, Matthews cutpoints, and NHANES cutpoints for assessing freeliving physical activity  
Millor et al.  An evaluation of the 30s chair stand test in older adults: frailty detection based on kinematic parameters from a single inertial unit  
HodtBillington et al.  Should trunk movement or footfall parameters quantify gait asymmetry in chronic stroke patients?  
US20140066816A1 (en)  Method, apparatus, and system for characterizing gait  
Dijkstra et al.  Detection of gait and postures using a miniaturized triaxial accelerometerbased system: accuracy in patients with mild to moderate Parkinson's disease  
Crouter et al.  Validity of the Actical for estimating freeliving physical activity  
Marschollek et al.  Assessing elderly persons' fall risk using spectral analysis on accelerometric dataa clinical evaluation study  
US20120092169A1 (en)  Method of assessing human fall risk using mobile systems  
US20110092860A1 (en)  System for clinical assessment of movement disorders  
Crouter et al.  Estimating physical activity in youth using a wrist accelerometer  
Kobsar et al.  Evaluation of agerelated differences in the stridetostride fluctuations, regularity and symmetry of gait using a waistmounted triaxial accelerometer  
s Calabr et al.  Validity of physical activity monitors for assessing lower intensity activity in adults 
Legal Events
Date  Code  Title  Description 

AS  Assignment 
Owner name: INTELGE CARE INNOVATIONS LLC, CALIFORNIA Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:GREENE, BARRY R.;O DONOVAN, ALAN D.;REEL/FRAME:027135/0502 Effective date: 20111024 

AS  Assignment 
Owner name: CARE INNOVATIONS, LLC, CALIFORNIA Free format text: CHANGE OF NAME;ASSIGNOR:INTELGE CARE INNOVATIONS LLC;REEL/FRAME:038780/0072 Effective date: 20160322 